Using artificial intelligence techniques to aid the search for new superconducting materials

For decades the search for new superconductors has relied on

the serendipity of material scientists to synthesize a new

material with superconducting proprieties. This is a very

slow and inefficient process. In recent years artificial

intelligence tools have been suggested to expedite the search.

We will describe several computational techniques we have

employed in our studies, which include: K-nearest neighbors

algorithm to perform classification of superconducting materials;

Random Forest regression to calculate superconducting critical

temperatures; Self-Organizing Maps and t-SNE to cluster and

visualize superconductors; and Generative Adversarial Networks

to predict new superconducting materials. These results will

be compared with the results of other similar studies. Our

most promising predictions will be discussed.